tree 0fda026feca26b163eb5a7f3150572f9c656e93a
parent 7d3987b0bd54bd3d381b94871c9873cf618154ce
author kunal642 <kunalkapoor642@gmail.com> 1569475082 +0530
committer akashrn5 <akashnilugal@gmail.com> 1579154652 +0530

[CARBONDATA-3592] Fix query on bloom in case of multiple data files in one segment

Problem:
1. Query on bloom datamap fails when there are multiple data files in one segment.
2. Query on bloom is giving wrong results in case of multiple carbondata files.

Solution:
1. Old pruned index files were cleared from the FilteredIndexSharedNames list. So further
pruning was not done on all the valid index files. Hence added a check to clear the index
files only in valid scenarios. Also handled the case where wrong blocklet id is passed while
creating the blocklet from relative blocklet id.
2. Make the partitions based on block path so that all the CarbonInputSplits in a MultiBlockSplit
are used for bloom reading. This means 1 task for 1 shard(unique block path).

This closes #3474